Registration of Large Motion Blurred Images
Technical Report,30 Sep 2014,29 Sep 2015
INDIAN INST OF TECH MADRAS MADRAS India
Pagination or Media Count:
Registration and change detection occupy an important place in aerial surveillance in which the goal is to mark regions of changes such as objects, buildings etc. over time. The challenge lies in handling the dynamics of the capturing system, for example, a drone. CMOS sensors, used in recent times, when employed in these cameras produce two types of distortions, namely the rolling shutter effect and motion blur. These are intertwined due to sequential row exposure mechanism. In this report, we propose a layered registration procedure to address the problem of detecting changes in 3D scenes affected by motion blur and rolling shutter artifacts. The proposed method has been tested on synthetic as well as real data.